NERRS banner

System-wide Monitoring Program

Synthesis of the Water Quality Data from 1995 to 2000
Chapter 1: Characteristics of Sampling

Introduction
A critical first step in evaluating and comparing data sets is to determine what, if any, sampling bias exists. Disproportionate sampling among sites, regions, seasons and years can complicate interpretation of observed differences in parameters at these levels. For example, several NERRs in the Northeast are not monitored during the winter months due to freezing temperatures and ice formation (Wenner et al. 2001). As a result, the annual frequency of occurrence of water temperatures <10°C at these sites are underestimated, and annual occurrence of supersaturation events may also be underestimated. Although it may not be appropriate to adjust values in these types of situations when comparing data sets, it is critical to document their occurrence and provide the necessary caveats for interpretation of results.

Extended deployment duration can lead to potential drift in dissolved oxygen (% sat) due to increased fouling of DO probes, which in turn represents a potential bias in determination of the frequency and duration of hypoxia events. Wenner et al. (2001) evaluated dissolved oxygen drift at NERR SWMP sites at various time intervals during the first 14 days post-deployment; however, the only criteria used to select these deployments was the total duration (³ 14 days). Because hypoxia at various time intervals was expressed as a percent of the time interval with DO < 28% sat, an additional criteria, the amount of DO data available at each time interval, should be specified to avoid bias. In order to determine if potential drift estimates reported by Wenner et al. (2001) were markedly different than estimates determined using the additional criteria, and to assess trends in hypoxia and supersaturation at varying time intervals over the entire data set, we re-evaluated potential drift at varying deployment time intervals.

Methods
Total deployments and deployment duration were evaluated to determine if sampling was similar among years, regions, and seasons. Deployments were assigned to each season according to the month in which the deployment was initiated. Seasons were defined as winter (January-March), spring (April-June), summer (July-September), and fall (October-December), and these definitions are used throughout this report. Data were plotted and graphically presented using MS Excel.

Total observations for each water quality parameter were determined on an annual and seasonal basis. Total observations were expressed as a percentage of the maximum number of potential 30-min observations for a given year or season. Maximum annual observations were 17,520, except during leap years (1996, 2000) when maximum annual observations increased to 17,568. Maximum observations were 4,368 in spring and 4,416 in summer and fall. Winter observations in leap years (1996, 2000) were 4,368, compared to 4,320 winter observations for other years.

Potential drift in dissolved oxygen (% saturation) due to fouling was evaluated by comparing the percent of time with hypoxia (<28% saturation) and supersaturation (>120% saturation) at various deployment intervals. Two hundred seventy deployments that began in July and August 1995-2000, were at least 14-d in duration, and contained at least 90% of DO (% sat) data for each deployment duration interval (1, 2, 4, 7, and 14-d) were selected for these analyses

Mean hypoxia and mean supersaturation were calculated from deployment-level percent of time observations for 1995-1996 (n=73), 1997-1998 (n=82), and 1999-2000 (n=115). Mean hypoxia and supersaturation for each deployment duration interval were plotted in MS Excel and a polynomial (quadratic) trend line fit to the data.

Regional evaluations of these data were also conducted to determine if differences existed among regions at sites monitored by the NERR SWMP. Mean and standard deviation for hypoxia and supersaturation were calculated from deployment-level percent of time observations between 1995-2000 for each of five geographic regions: West Coast (n=70), Northeast/Great Lakes (n=75), Mid-Atlantic (n=44), Southeast (n=36), and the Gulf of Mexico/Caribbean (n=45). Mean and standard deviation of the percent of time with hypoxia and supersaturation were plotted for each deployment duration interval.

Results and Discussion
Water quality observations collected by the NERR SWMP between 1995-2000 total 3.18 million records from 55 sites and 22 Reserves. Five sites not included in the previous synthesis report that contained partial or complete 1996-1998 data are included in the current study because these sites were still actively participating in the NERR SWMP at the end of 2000. Six additional sites were added to the NERR SWMP in 1999-2000, three of which replaced sites monitored by these Reserves during 1996-1998 and three of which were added as the third site at their respective Reserves in 1999 and 2000.

During the first six years of the NERR SWMP (1995-2000), a total of 4,135 YSI deployments conducted at 55 monitoring sites were included in analyses for this report. Due to the progressive implementation of the SWMP at NERR sites, total deployments in 1995 (n=366) were substantially lower than the total number of annual deployments for the other four years. Similar numbers of total deployments were conducted in 1996 (n=626) and 1997 (n=691). Similar numbers of deployments were also conducted in 1998 (n=796) and 1999 (n=783); however, these deployments represented an increase of approximately 15% from 1996-1997 levels. Total deployments in 2000 (n=873) were the most observed for a single year between 1995-2000 and represented an increase of approximately 10% from 1998-1999 levels.

Deployments conducted at sites in the Southeast (n=988) and Mid-Atlantic (n=956) regions collectively accounted for approximately half of the total deployments, and the number of annual deployments per site in each of these regions averaged 19.3 and 18.0, respectively (Figure 1). Similar numbers of deployments were conducted at NERR sites in the Northeast/Great Lakes region (n=777) and Gulf of Mexico/Caribbean region (n=778); however, relative sampling effort was drastically different at sites within these two regions (11.4 vs. 17.7 annual deployments/site). Substantially fewer deployments were conducted at NERR sites on the West Coast (n=636) where relative sampling effort averaged 12.7 annual deployments per site in this region.

Annual sampling effort among regions, NERR SWMP 1995-2000.

Seasonal variation in deployment frequency (Figure 2) and duration (Figure 3) were apparent. Maximum deployment frequency (3.5 to 6.1 deployments per site) was typical for all regions during summer while minimum deployment frequency (0.6 to 4.3 deployments per site) was typical during winter. Maximum deployment duration (15.2-24.7 d) was typically observed for all regions in winter while minimum deployment duration (12.3-19.3 d) was typically observed in summer. Seasonal deployment (frequency and duration) trends were particularly noticeable for NERRs in the Northeast/Great Lakes, Mid-Atlantic regions, and West Coast regions.

Seasonal deployment trends resulted in variable data collection (Appendices 1-9). Many reserves collected more water quality observations in summer than in winter. This trend was especially pronounced for Great Bay, Hudson River, Old Woman Creek, Waquoit Bay, Wells (Head of Tide), and Chesapeake Bay MD (Appendices 4-9). Inter-annual variability was irregular for a number of sites (Appendices 1-3). Inter-annual variability was most noticeable in years when monitoring sites were either initiated or terminated. As such, the percent of monitoring sites that collected at least 50% of the maximum number of annual observations steadily improved for all water quality parameters between 1995-2000; however, even in year 2000, 7-14 sites (13-27%) did not collect at least 50% of annual data for most water quality parameters (Figure 4).


Seasonal sampling effort among regions, NERR SWMP 1995-2000.


Mean regional and seasonal deployment duration, NERR SWMP 1995-2000.


Percent of NERR sites collecting at least 50% of annual water quality data for each parameter by year.

Mean percent of time with hypoxia vs. deployment duration (1, 2, 4, 7, & 14-d intervals) was similar among year groups (quadratic increase, R2=0.95-0.98, Figure 5). Incidentally, hypoxia at varying post-deployment intervals also progressively increased between biennial groupings. Mean percent of supersaturation vs. deployment duration (1, 2, 4, 7 & 14-d intervals) was also similar among year groups (quadratic decay, R2=0.98-0.99, Figure 6) and decreased between biennial groupings.

Regional differences in hypoxia and supersaturation at NERR sites were apparent; however, within each region, the overall trends of quadratic increase (hypoxia, Figure 7) and quadratic decrease (supersaturation, Figure 8) with increasing post-deployment duration were observed. Mean percent of hypoxia (and corresponding standard deviation) was lowest at NERR sites located in the Northeast/Great Lakes and Southeast regions. The occurrence of low hypoxia at sites in the Southeast region was perplexing, but likely represents numerous sites located in relatively pristine water bodies. Mean percent of supersaturation was greatest at NERR sites located on the West Coast. Mean percent of supersaturation was lowest (and remarkably similar) for NERR sites located in the Northeast/Great Lakes and Mid-Atlantic regions.

Given these observations, hypoxia and supersaturation analyses were again based only on the first 48-hours post-deployment, consistent with the Wenner et al. (2001) report. (For more information see PDF)